In speech communications on an IP network, to realize network traffic control and multicast communication on the network, a scalable function, which enables a receiving apparatus to acquire decoded speech of certain quality even from part of encoded data, is anticipated.
In scalable coding (scalable speech coding) having this scalable function, by encoding an input speech signal into layers, encoded data with a plurality of layers including the lower layer to higher layers, are generated and transmitted. The receiving apparatus acquires decoded speech using encoded data with the lower layer to an arbitrary higher layer and thereupon acquires a decoded signal in varying quality, thereby decoding the speech in higher quality by decoding higher layers. Here, enhancement layer encoded data is directed to improving quality of the core layer.
By the way, when frame loss occurs in a channel, there is a technique of performing frame erasure concealment by extrapolating parameters received earlier in a speech decoding apparatus. However, for example, it is difficult to estimate a signal of speech onset using only the parameters received earlier. Consequently, it is not practical to realize robustness to packet loss using only the method of extrapolation-based concealment.
Therefore, besides extrapolation, there is a technique of adding in advance redundancy information for concealment processing upon transmission (see Patent Documents 1 and 2). By separately transmitting encoded data for concealment generated from this concealment information, it is possible to enhance error robustness.
Patent Document 1 discloses a technique of encoding the current frame by the first coding method, and, using its decoded signal, encoding a future signal by a second coding method (sub-codec), and outputting both encoded data at the same time. In this case, if the first encoded data is lost, high error robustness is realized by performing concealment using the second encoded data received earlier.
Patent Document 2 discloses a technique of encoding the current frame by the first coding method, extracting and encoding periodicity information such as the pitch of the future frame for packet loss concealment, and transmitting both data at the same time. As in Patent Document 1, if the encoded data of the current frame is lost, high error robustness is realized by performing concealment using the encoded data for concealment, which is received earlier.
Patent Documents 1 and 2 disclose using encoded data from a sub-codec which targets other periods than the current frame as encoded data for concealment, and transmitting this encoded data and the encoded data of the current frame by the first coding scheme at the same time. By this means, even when the encoded data of the current frame is lost, error robustness is emphasized by performing concealment using the supplementary information.    Patent Document 1: Japanese Patent Application Laid-Open No. 2002-221994    Patent Document 2: Japanese Patent Application Laid-Open No. 2002-268696